up-cli 0.2.0__tar.gz → 0.5.0__tar.gz

up_cli-0.5.0/.claude/context_budget.json.migrated

@@ -0,0 +1,12 @@
+{
+  "budget": 100000,
+  "total_tokens": 0,
+  "remaining_tokens": 100000,
+  "usage_percent": 0.0,
+  "status": "OK",
+  "warning_threshold": 0.8,
+  "critical_threshold": 0.9,
+  "session_start": "2026-02-04T18:17:59.380676",
+  "entry_count": 0,
+  "entries": []
+}

{up_cli-0.2.0 → up_cli-0.5.0}/.claude/settings.local.json

@@ -10,7 +10,8 @@
       "Bash(pip show:*)",
       "Bash(git add:*)",
       "Bash(git commit:*)",
-      "WebFetch(domain:img.iami.xyz)"
+      "WebFetch(domain:img.iami.xyz)",
+      "Bash(python:*)"
     ]
   }
 }

up_cli-0.5.0/.claude/skills/docs-system/SKILL.md

@@ -0,0 +1,22 @@
+---
+name: docs
+description: Documentation system with standards
+user-invocable: true
+allowed-tools: Read, Write, Edit, Glob
+---
+
+# Docs Skill
+
+## Commands
+
+- `/docs new [type]` - Create document
+- `/docs status` - Show status
+
+## Types
+
+| Type | Folder |
+|------|--------|
+| feature | features/ |
+| arch | architecture/ |
+| changelog | changelog/ |
+| guide | guides/ |

up_cli-0.5.0/.claude/skills/learning-system/SKILL.md

@@ -0,0 +1,204 @@
+---
+name: learn
+description: Research and create improvement plans
+user-invocable: true
+allowed-tools: Read, Write, Bash, WebFetch, WebSearch, Glob, Grep
+version: "1.0.0"
+min-claude-version: "2024.01"
+---
+
+# Learning System
+
+Research best practices from open source projects and blogs, then create actionable improvement plans.
+
+## Workflow
+
+```
+RESEARCH → ANALYZE → COMPARE → PLAN → IMPLEMENT
+```
+
+## Commands
+
+| Command | Description |
+|---------|-------------|
+| `/learn auto` | Auto-analyze project and generate insights |
+| `/learn research [topic]` | Research a specific topic |
+| `/learn analyze` | Analyze all research files |
+| `/learn compare` | Compare findings with current code |
+| `/learn plan` | Generate improvement PRD |
+| `/learn full` | Run complete pipeline |
+
+## Phase 1: RESEARCH
+
+### Default Sources
+
+Read `sources.json` for configured sources. Research from:
+
+**Open Source Projects:**
+- semgrep/semgrep - Pattern-based code analysis
+- github/codeql - Query-based analysis
+- High-quality projects in your tech stack
+
+**Blogs & Documentation:**
+- Official documentation for your frameworks
+- Engineering blogs from top companies
+
+### Actions
+
+1. Use `WebSearch` to find recent articles/releases
+2. Use `WebFetch` to read project READMEs and docs
+3. Extract architecture patterns
+4. Note innovative approaches
+
+### Output
+
+Save findings to `research/YYYY-MM-DD-topic.md`:
+
+```markdown
+# Research: [Topic Name]
+
+**Created**: YYYY-MM-DD
+**Status**: 📋 Reference
+**Source**: [Project/Blog Name]
+
+---
+
+## Key Findings
+
+1. Finding 1
+2. Finding 2
+
+## Applicable Patterns
+
+- Pattern A: Description
+- Pattern B: Description
+
+## Code Examples
+
+```language
+// example code
+```
+```
+
+---
+
+## Phase 2: ANALYZE
+
+Read all research files and extract:
+
+1. **Design Patterns** - Architectural approaches
+2. **Key Innovations** - Novel techniques
+3. **Best Practices** - Industry standards
+4. **Reusable Components** - Code/concepts to adopt
+
+### Output
+
+Update `insights/patterns.md` with extracted patterns.
+
+---
+
+## Phase 3: COMPARE
+
+Compare insights with current implementation:
+
+1. Read project's main documentation (CLAUDE.md, README.md)
+2. Scan source code for current patterns
+3. Identify gaps and opportunities
+
+### Output
+
+Update `insights/gap-analysis.md` with findings.
+
+---
+
+## Phase 4: PLAN
+
+Create improvement plan as PRD:
+
+1. Prioritize gaps by impact/effort
+2. Create user stories with acceptance criteria
+3. Order by dependencies
+4. Save as `prd.json`
+
+### PRD Schema
+
+```json
+{
+  "project": "Project Name",
+  "branchName": "feature/improvements",
+  "description": "Brief description",
+  "userStories": [
+    {
+      "id": "US-001",
+      "title": "Story title",
+      "description": "Description",
+      "acceptanceCriteria": ["Criterion 1", "Criterion 2"],
+      "priority": 1,
+      "effort": "low|medium|high",
+      "passes": false
+    }
+  ]
+}
+```
+
+---
+
+## Phase 5: IMPLEMENT
+
+Hand off to Product Loop:
+
+```
+/product-loop
+```
+
+The product loop will:
+1. Read `prd.json` generated by learning system
+2. Execute tasks with circuit breaker protection
+3. Checkpoint before risky changes
+4. Track progress
+
+---
+
+## AUTO Mode
+
+When `/learn auto` is invoked:
+
+1. Scan the codebase to detect technologies and patterns
+2. Identify improvement areas automatically
+3. Generate research topics based on findings
+4. Research each topic (limit: 3 topics)
+5. Continue with analyze → compare → plan
+
+### Auto Mode Output
+
+```
+Project Profile:
+- Languages: Python, TypeScript
+- Frameworks: FastAPI, React
+- Patterns: Repository Pattern, React Hooks
+- Improvements: add-caching, optimize-queries
+- Topics: FastAPI caching best practices, React performance
+```
+
+---
+
+## State Files
+
+| File | Purpose |
+|------|---------|
+| `sources.json` | Research sources config |
+| `research/*.md` | Raw research notes |
+| `insights/patterns.md` | Extracted patterns |
+| `insights/gap-analysis.md` | Gap analysis |
+| `prd.json` | Generated improvement plan |
+
+---
+
+## Context Budget
+
+This skill is context-aware:
+- Estimates token usage before reading files
+- Warns when approaching budget limits
+- Suggests summarization when needed
+
+Check budget: Read `.claude/context_budget.json`

up_cli-0.5.0/.claude/skills/learning-system/file_learnings/2026-02-04_git_for_ai_vibe_coding.json

@@ -0,0 +1,41 @@
+{
+  "source_file": "Git for AI Vibe Coding.txt",
+  "source_path": "/Users/mour/AI/up-cli/Git for AI Vibe Coding.txt",
+  "file_type": ".txt",
+  "key_concepts": [
+    "!/bin/bash",
+    "Save as test_regression.sh and make executable (chmod +x)",
+    "1. Build the project. If the build fails, we can't test.",
+    "We return 125 to tell git bisect to 'skip' this commit.",
+    "2. Run the specific test case that reproduces the bug.",
+    "Do NOT run the full suite; it's too slow. Focus on the regression.",
+    "3. Return exit code to git bisect",
+    "0 = Good (Bug not present)",
+    "1 = Bad (Bug present)",
+    ".cursorrules",
+    "Add to.zshrc or.bashrc"
+  ],
+  "patterns_found": [
+    "Design pattern mentioned",
+    "Best practice documented",
+    "Architecture concept",
+    "Workflow described",
+    "Convention defined",
+    "Standard referenced"
+  ],
+  "best_practices": [
+    "Constraint Example: \"Before applying changes to more than 3 files, propose a plan.\"",
+    "Constraint Example: \"Never delete comments marked // CRITICAL.\"",
+    "The Cost: This can lead to \"commit explosion.\" A feature that takes a human 3 commits might take Aid",
+    "Create branch feat/login-page from main.",
+    "Run git worktree add.worktrees/login-page feat/login-page.",
+    "Copy necessary .env files or context configs to the new directory.",
+    "Open a new terminal tab/pane in that directory.",
+    "git bisect start",
+    "git bisect bad (Current broken state)",
+    "git bisect good <commit-hash-from-yesterday>"
+  ],
+  "code_snippets": [],
+  "action_items": [],
+  "ai_analysis": "I'll analyze this document on Git strategies for AI-driven \"vibe coding\" development.\n\n## Key Concepts\n\n1. **Vibe Coding Paradigm** - Development where intent orchestration replaces manual coding; developer acts as architect directing AI builders\n\n2. **Pure vs Responsible Vibe Coding** - Pure is exploratory/rapid prototyping; Responsible maintains code ownership with rigorous review\n\n3. **The \"70% Problem\"** - AI gets you 70% there instantly, but final 30% of integration/debugging takes exponentially longer\n\n4. **\"Paint Sprayer\" Effect** - AI generates code as \"black boxes\" vs the \"white box\" understanding of manual coding\n\n5. **Version Loop Tightening** - Traditional \"Code \u2192 Test \u2192 Commit\" becomes \"Prompt \u2192 Generate \u2192 Visual Diff \u2192 Reset/Commit\"\n\n6. **Git as \"Ripcord\"** - Commit before every significant prompt; `git reset --hard` immediately on flawed generation\n\n7. **Provenance Obscurity** - `git blame` shows who prompted, but the reasoning/prompt context is lost\n\n8. **Contextual Overwrite Risk** - AI may delete critical edge-case handling outside its context window\n\n## Patterns\n\n| Pattern | Description |\n|---------|-------------|\n| **Commit-Before-Prompt** | Always commit clean state before AI generation |\n| **Hard Reset Recovery** | Reset immediately on bad generation vs \"prompting out\" |\n| **Git Worktrees** | Parallel development with multiple AI agents |\n| **Specification-Driven** | Tools like Copilot Workspace use specs before code |\n| **Composer Model** | IDE-native AI editing multiple files simultaneously |\n\n## Best Practices\n\n1. **Never \"prompt your way out\"** of bad generation - reset and refine the prompt instead\n2. **Use `.cursorrules`** (or equivalent) as a gatekeeper for AI behavior\n3. **Review every diff** in responsible mode - treat AI as pair programmer, not autonomous agent\n4. **Commit atomically** - small, logical commits for easier bisect debugging\n5. **Preserve prompt context** - document the \"why\" alongside commits for future debugging\n6. **Avoid \"Apply All\" blindly** - scrutinize multi-file changes before accepting\n7. **Use worktrees for parallel agents** - prevents agents from overwriting each other\n\n## Implementation Ideas\n\n### For up-cli\n\n1. **Add pre-prompt checkpoint command** - Auto-commit or stash before AI operations\n2. **Prompt logging** - Store prompts alongside commits for provenance\n3. **Worktree orchestration** - Manage parallel agent branches with automatic merge conflict detection\n4. **\"Doom loop\" detection** - Warn when multiple failed prompts suggest need for reset\n5. **Diff review workflow** - Integrate visual diff step before accepting AI changes\n6. **Context budget awareness** - Track AI context window usage to prevent contextual overwrite\n\n### Workflow Integration\n\n```\n# Proposed vibe coding safety loop\n1. git commit -m \"checkpoint: before AI prompt\"\n2. Run AI generation\n3. Review diff (mandatory)\n4. Accept \u2192 commit with prompt context\n   OR Reject \u2192 git reset --hard HEAD\n```"
+}

up_cli-0.5.0/.claude/skills/learning-system/file_learnings/2026-02-04_git_for_ai_vibe_coding.md

@@ -0,0 +1,114 @@
+# Learnings from: Git for AI Vibe Coding.txt
+
+**Analyzed**: 2026-02-04
+**Source**: `/Users/mour/AI/up-cli/Git for AI Vibe Coding.txt`
+**Type**: .txt
+**Method**: AI-powered
+
+## AI Analysis
+
+I'll analyze this document on Git strategies for AI-driven "vibe coding" development.
+
+## Key Concepts
+
+1. **Vibe Coding Paradigm** - Development where intent orchestration replaces manual coding; developer acts as architect directing AI builders
+
+2. **Pure vs Responsible Vibe Coding** - Pure is exploratory/rapid prototyping; Responsible maintains code ownership with rigorous review
+
+3. **The "70% Problem"** - AI gets you 70% there instantly, but final 30% of integration/debugging takes exponentially longer
+
+4. **"Paint Sprayer" Effect** - AI generates code as "black boxes" vs the "white box" understanding of manual coding
+
+5. **Version Loop Tightening** - Traditional "Code → Test → Commit" becomes "Prompt → Generate → Visual Diff → Reset/Commit"
+
+6. **Git as "Ripcord"** - Commit before every significant prompt; `git reset --hard` immediately on flawed generation
+
+7. **Provenance Obscurity** - `git blame` shows who prompted, but the reasoning/prompt context is lost
+
+8. **Contextual Overwrite Risk** - AI may delete critical edge-case handling outside its context window
+
+## Patterns
+
+| Pattern | Description |
+|---------|-------------|
+| **Commit-Before-Prompt** | Always commit clean state before AI generation |
+| **Hard Reset Recovery** | Reset immediately on bad generation vs "prompting out" |
+| **Git Worktrees** | Parallel development with multiple AI agents |
+| **Specification-Driven** | Tools like Copilot Workspace use specs before code |
+| **Composer Model** | IDE-native AI editing multiple files simultaneously |
+
+## Best Practices
+
+1. **Never "prompt your way out"** of bad generation - reset and refine the prompt instead
+2. **Use `.cursorrules`** (or equivalent) as a gatekeeper for AI behavior
+3. **Review every diff** in responsible mode - treat AI as pair programmer, not autonomous agent
+4. **Commit atomically** - small, logical commits for easier bisect debugging
+5. **Preserve prompt context** - document the "why" alongside commits for future debugging
+6. **Avoid "Apply All" blindly** - scrutinize multi-file changes before accepting
+7. **Use worktrees for parallel agents** - prevents agents from overwriting each other
+
+## Implementation Ideas
+
+### For up-cli
+
+1. **Add pre-prompt checkpoint command** - Auto-commit or stash before AI operations
+2. **Prompt logging** - Store prompts alongside commits for provenance
+3. **Worktree orchestration** - Manage parallel agent branches with automatic merge conflict detection
+4. **"Doom loop" detection** - Warn when multiple failed prompts suggest need for reset
+5. **Diff review workflow** - Integrate visual diff step before accepting AI changes
+6. **Context budget awareness** - Track AI context window usage to prevent contextual overwrite
+
+### Workflow Integration
+
+```
+# Proposed vibe coding safety loop
+1. git commit -m "checkpoint: before AI prompt"
+2. Run AI generation
+3. Review diff (mandatory)
+4. Accept → commit with prompt context
+   OR Reject → git reset --hard HEAD
+```
+
+---
+
+## Key Concepts (Basic Extraction)
+
+- !/bin/bash
+- Save as test_regression.sh and make executable (chmod +x)
+- 1. Build the project. If the build fails, we can't test.
+- We return 125 to tell git bisect to 'skip' this commit.
+- 2. Run the specific test case that reproduces the bug.
+- Do NOT run the full suite; it's too slow. Focus on the regression.
+- 3. Return exit code to git bisect
+- 0 = Good (Bug not present)
+- 1 = Bad (Bug present)
+- .cursorrules
+- Add to.zshrc or.bashrc
+
+## Patterns Found
+
+- [ ] Design pattern mentioned
+- [ ] Best practice documented
+- [ ] Architecture concept
+- [ ] Workflow described
+- [ ] Convention defined
+- [ ] Standard referenced
+
+## Best Practices
+
+- [ ] Constraint Example: "Before applying changes to more than 3 files, propose a plan."
+- [ ] Constraint Example: "Never delete comments marked // CRITICAL."
+- [ ] The Cost: This can lead to "commit explosion." A feature that takes a human 3 commits might take Aid
+- [ ] Create branch feat/login-page from main.
+- [ ] Run git worktree add.worktrees/login-page feat/login-page.
+- [ ] Copy necessary .env files or context configs to the new directory.
+- [ ] Open a new terminal tab/pane in that directory.
+- [ ] git bisect start
+- [ ] git bisect bad (Current broken state)
+- [ ] git bisect good <commit-hash-from-yesterday>
+
+## Action Items
+
+- [ ] Review insights and apply to project
+- [ ] Run `up learn analyze` to process all learnings
+- [ ] Run `up learn plan` to generate improvement PRD

up_cli-0.5.0/.claude/skills/learning-system/insights/2026-02-04_git_for_ai_vibe_coding_insights.md

@@ -0,0 +1,69 @@
+# Insights: 2026-02-04_git_for_ai_vibe_coding.md
+
+**Analyzed**: 2026-02-04
+**Source**: `/Users/mour/AI/up-cli/.claude/skills/learning-system/file_learnings/2026-02-04_git_for_ai_vibe_coding.md`
+**Method**: claude CLI
+
+---
+
+I'll analyze this research document and extract the key information.
+
+## Analysis: Git for AI Vibe Coding
+
+### 1. Key Patterns
+
+| Pattern | Description |
+|---------|-------------|
+| **Commit-Before-Prompt** | Always commit clean state before AI generation |
+| **Hard Reset Recovery** | Reset immediately on bad generation vs "prompting out" |
+| **Git Worktrees** | Parallel development with multiple AI agents in separate directories |
+| **Specification-Driven** | Use specs/rules files before code generation |
+| **Composer Model** | IDE-native AI editing multiple files simultaneously |
+| **Version Loop Tightening** | "Prompt → Generate → Visual Diff → Reset/Commit" replaces traditional cycle |
+
+### 2. Best Practices
+
+1. **Never "prompt your way out"** - Reset and refine prompt instead of iterating on bad code
+2. **Use `.cursorrules`/`CLAUDE.md`** - Gatekeeper files for AI behavior constraints
+3. **Review every diff** - Treat AI as pair programmer, not autonomous agent
+4. **Commit atomically** - Small, logical commits for easier `git bisect` debugging
+5. **Preserve prompt context** - Document the "why" alongside commits
+6. **Avoid "Apply All" blindly** - Scrutinize multi-file changes
+7. **Use worktrees for parallel agents** - Prevents agents from overwriting each other
+
+### 3. Gaps Identified
+
+| Gap | Impact |
+|-----|--------|
+| **Provenance Obscurity** | `git blame` shows who prompted, but reasoning/prompt context is lost |
+| **Contextual Overwrite Risk** | AI may delete critical edge-case handling outside its context window |
+| **The "70% Problem"** | Final 30% of integration/debugging takes exponentially longer |
+| **"Paint Sprayer" Effect** | AI generates "black boxes" vs "white box" understanding |
+| **Commit Explosion** | Features that take humans 3 commits may take 30+ with AI |
+
+### 4. Action Items
+
+**Immediate (for up-cli)**:
+- [ ] Add pre-prompt checkpoint command (auto-commit/stash before AI ops)
+- [ ] Implement "doom loop" detection (warn after multiple failed prompts)
+- [ ] Add context budget tracking to prevent contextual overwrite
+
+**Short-term**:
+- [ ] Prompt logging - store prompts alongside commits for provenance
+- [ ] Diff review workflow - integrate mandatory visual diff step
+- [ ] Worktree orchestration - manage parallel agent branches
+
+**Workflow to Implement**:
+```bash
+# Vibe coding safety loop
+1. git commit -m "checkpoint: before AI prompt"
+2. Run AI generation
+3. Review diff (mandatory)
+4. Accept → commit with prompt context
+   OR Reject → git reset --hard HEAD
+```
+
+**Constraint Examples for CLAUDE.md**:
+- "Before applying changes to more than 3 files, propose a plan"
+- "Never delete comments marked `// CRITICAL`"
+

up_cli-0.5.0/.claude/skills/learning-system/insights/2026-02-05-implementation-review.md

@@ -0,0 +1,119 @@
+# Implementation Review Analysis
+
+**Created**: 2026-02-05
+**Status**: 🔄 Active
+**Source**: Self-analysis of up-cli codebase
+
+---
+
+## Summary
+
+Analysis of the up-cli project after completing all 26 PRD tasks revealed several gaps between "marked complete" and "actually complete."
+
+## Key Findings
+
+### 1. Dead Code (Critical)
+
+**File**: `src/up/commands/learn.py` (1741 lines)
+
+- Task A-001 "Split learn.py into submodules" marked complete
+- New modules created in `src/up/learn/`
+- CLI updated to use new modules
+- **OLD FILE NEVER DELETED**
+
+**Impact**: 1741 lines of dead code, confusing for maintainers
+
+### 2. Zero Test Coverage (High)
+
+| Category | Files | Tests | Coverage |
+|----------|-------|-------|----------|
+| Core (state, checkpoint, provenance) | 3 | 0 | 0% |
+| Commands (start, vibe, agent, etc.) | 12 | 0 | 0% |
+| Learn modules | 5 | 0 | 0% |
+| Git utilities | 2 | 0 | 0% |
+
+**Impact**: No confidence in code correctness, regressions likely
+
+### 3. Template-Code Drift (Medium)
+
+Templates in `src/up/templates/` contain:
+- Full `circuit_breaker.py` implementation
+- Full `state_manager.py` implementation
+- Both diverge from actual `src/up/core/` implementations
+
+**Impact**: New projects get outdated code patterns
+
+### 4. Large Monolithic Files Remain (Medium)
+
+| File | Lines | Should Be |
+|------|-------|-----------|
+| `commands/start.py` | 1077 | < 400 (split by phase) |
+| `memory.py` | 1096 | < 400 (split by function) |
+
+## Patterns Identified
+
+### Anti-Pattern: "Mark Complete Without Verification"
+
+Task completion criteria should include:
+1. ✅ Code written
+2. ❌ Old code removed (missing)
+3. ❌ Tests written (missing)
+4. ❌ Templates updated (missing)
+
+### Anti-Pattern: "Reference Implementation in Templates"
+
+Templates should either:
+- Import from actual modules, OR
+- Contain minimal stubs with "see src/up/core/" comments
+
+Not full implementations that drift.
+
+## Recommendations
+
+### Immediate (Sprint 8: Cleanup)
+
+1. **Delete dead code**
+   - Remove `src/up/commands/learn.py`
+   
+2. **Add core tests**
+   - `tests/test_state.py` - State management
+   - `tests/test_checkpoint.py` - Checkpoint operations
+   - `tests/test_provenance.py` - Provenance tracking
+
+3. **Fix template drift**
+   - Update `templates/loop/__init__.py` to NOT include full implementations
+   - Add note pointing to `src/up/core/`
+
+### Short-term (Sprint 9: Quality Gates)
+
+4. **Add test requirement to PRD process**
+   - Task not complete without test
+   - Add `tests_required: true` to PRD schema
+
+5. **Add coverage tracking**
+   - Configure pytest-cov
+   - Add badge to README
+   - Minimum 60% coverage gate
+
+### Medium-term (Sprint 10: Modularization)
+
+6. **Split start.py**
+   - `start/loop.py` - Main loop logic
+   - `start/phases.py` - Phase implementations
+   - `start/verification.py` - Test/lint verification
+
+7. **Split memory.py**
+   - `memory/storage.py` - ChromaDB operations
+   - `memory/search.py` - Semantic search
+   - `memory/index.py` - File indexing
+
+---
+
+## Metrics
+
+| Metric | Current | Target |
+|--------|---------|--------|
+| Dead code files | 1 | 0 |
+| Test coverage | ~2% | 60% |
+| Max file size | 1741 | 400 |
+| Template drift items | 2 | 0 |